With the help of Random Forest, Billerud has built a modern data platform that collects and refines massive volumes of process and business data – using real-time capabilities in selected areas. The result is a more flexible, unified and future-proof environment that enables faster decision-making, higher quality and a stronger data-driven culture across the entire organization.
The challenge
As one of the world's leading producers of packaging materials, Billerud faces a huge amount of data every day. The company has manufacturing operations in both Europe and North America, producing packaging board, liquid packaging board, sack and kraft paper around the clock. Each mill is equipped with tens of thousands of sensors, generating over 15 billion data points per year – just from the production processes.
Historically, Billerud has used a traditional BI architecture with ETL flows, data warehouses in SQL Server and reporting via Power BI. The model worked, but had clear limitations. The platform was batch-based, making it difficult to handle real-time data and process information. New source systems in the cloud also added complexity and BI management often became a bottleneck when the business needed new data sources or reports quickly.
"Like many others, we realized that our existing platform was no longer sufficient. We needed a modern, unified and scalable environment where the entire organization could store, analyze and share data – without being dependent on the central BI management," says Henrik Johansson, Data Architect at Billerud and project manager for the platform.
In addition to the technical challenges, there was also an organizational need: to give more parts of the business the opportunity to develop and deploy data flows themselves, without compromising structure, quality and security.
The solution
Together with Random Forest, Billerud defined a new target architecture and built the modern data platform BLADE (Billerud Large Data Environment). At its core is Snowflake, which enables data to be stored, transformed and analyzed regardless of type or source. The platform is designed around a clear domain model, where each domain has its own database but follows the same framework for orchestration, structure and governance.
"We chose to put as much as possible in SQL directly in Snowflake. This lowers the threshold for business teams who want to develop in the platform because you don't need to know advanced coding, just SQL. It allows more people to contribute," says Johansson.
Random Forest was responsible for the implementation and architecture, with a strong focus on making the platform as code- and metadata-driven as possible. By updating metadata alone, Billerud can now rapidly add new data sources, pipelines and use cases.
The architecture is based on three layers – bronze, silver and gold – where data is gradually cleansed and harmonized. At the top is a domain interface that acts as a data contract between different parts of the organization and ensures a "single source of truth".
A central component is also the control hub developed together with Random Forest, where the entire framework is developed and distributed to all domains. This allows teams to work independently but still within a common standard.
“We are a traditional process industry with relatively limited IT resources, so it was important that the platform would be easy to manage and scale. By consolidating everything into a single toolset, we’ve reduced complexity and gained better visibility,” Johansson explains.
One concrete example of how the platform is used is in the handling of quality data. Previously, data was scattered, difficult to maintain and heavy to load in Power BI. After the introduction of BLADE, the domain owner can now build and maintain the entire data pipeline – from ingestion to reporting – directly within the platform. Heavy transformations are handled in Snowflake rather than in Power BI, resulting in faster, more reliable reports and an improved user experience.
Business benefits
The introduction of BLADE has given Billerud a completely new way of working with data. The platform is now in full operation and the results are clearly visible in the business:
- Shorter lead times from idea to finished report or analysis.
- Increased availability and performance in Power BI reports.
- Autonomous business teams that can develop and deploy their own data solutions.
- Unified and traceable data management that reduces duplication and sources of error.
- Scalable architecture that can handle both batch and real-time data flows.
For quality monitoring, this has meant that data on defects, material properties and production campaigns are now managed in a common structure. In a single report, you can now analyze billions of data points – something that was previously technically impossible.
“There is a completely different level of stability and transparency now. We can monitor production data and quality outcomes in a whole new way, share insights with customers and act quickly when something deviates,” Johansson notes.
The project has also contributed to a clearer data-driven way of working throughout Billerud. By starting with a limited MVP case and then building on it step by step, the team was able to quickly demonstrate value and build trust internally.
"One key lesson is not to wait for perfection – good enough is good enough. When a solution works, we move on to the next. That's how you build something sustainable over time," Johansson concludes.
Listen to Henrik Johansson talk about the collaboration at the Data Day 2025:
